Research On Optimization Of Multicast Models And Algorithms In P2P Media Streaming Network

Traditional Streaming Media Technology is based on client-server. A lot of problems appear along with the dramatic development of user numbers, particularly when a great amount of users are online simultaneously. In such cases, the performance of server drops off sharply thus cannot meet user's demand. As a new application in fields of network, P2P network model enjoys great potential in aspects like availability, expansibility and fault tolerance and could make full use of variety kinds of client resources. Regard P2P streaming media as research object, our research starts from its system frame, and further investigates the optimization algorithm of inter and intra multicast tree as well as the construction model of multicast tree.A hierarchical P2P streaming media system(HPSS) frame model is proposed to support high video broadcast rate. This system adopts multilevel cluster topological structure network, take cluster strategy according to latency and bandwidth. All nodes are clustered in a cluster of balanced bandwidth thus to get the maximum speed of streaming media. In addition, design algorithm of nodes' leaving and joining cluster to guarantee that HPSS can provide consistently stable service of high play rate. Through simulation experiment, HPSS make full use of bandwidth of nodes and can get speed close to the limitation speed of streaming media. This model maximizes the streaming media speed by the efficient maintaining of hierarchical structure thus solve the problem of high play rate in streaming media.A kind of hierarchical application multicast model based on priority (PH-ALM) is proposed. This model considers bandwidth and latency comprehensively and come up with the concept of priority. On the basis of maintaining original hierarchical ALM model, construct local ALM tree based on the standard of priority. Priority has strict mathematical definition and calculation method, fully considers local bandwidth and latency among nodes. The local ALM tree constructed by the standard priority has local optimality. Meanwhile, this model designs highly efficient algorithm for nodes' leaving, joining and position adjusting. Moreover, it makes ALM tree adjust positions regularly, balances the load distribution of data retransmission, adapts heterogeneous bandwidth of hosts and reinforces the flexibility and ALM structure. The results of simulation experiments prove that this model can well adapt the networks with great different node ability, and the cost of maintaining is relatively small, the optimization of structure of ALM tree in cluster limit in local which guarantees the stability of the system.An optimization of inter overlay algorithm based on history information reusing (HIR) is proposed. This algorithm does inter overlay optimization based on intra overlay optimization. Make each node keep several connected tables of both inter and intra tree. The content is the statistics of the history information of the inter-and intra-tree's optimal path this node takes part in. In the aspect of the investigation of optimal path, this algorithm improves the original reverse detection algorithm, overcome the problem of randomness when choosing neighbour, chooses the nodes with greatest SUM as neighbour nodes according to the record of total numbers of each nodes chosen as one of the node of optimal path, highly improves the hit rate of the choose of optimal path, makes the optimization of inter overlay more complete. The results of experiments show that HIR improves the efficiency of optimization of inter overlay, the use ratio of system resource and the overall system performance.A dynamic tree algorithm based on service window is proposed. This algorithm introduces service and steaming media server keeps a service window for each on-line video, the size of window determines the amount of nodes which could receive the video objects directly from the service window. Moreover, adjust the window size according to the history state of service window during certain period. This algorithm also sets different strategies for each nodes' leaving, joining and adjusting thus to balance the load and guarantee the robustness. The results of simulations show that the model could well balance the load among different streaming video. At the meantime of emphasizing high bandwidth, the algorithm also introduces reserve mechanism and has high bandwidth use ratio.Simulate experiments of the method above respectively have done to analyze their pros and cons from the side of experiments and compare the results with traditional policies. The results show that we get improvement in the average latency between ends, availability of link's resource, the load balance of data transmitting, the recovery rate of disable multicast tree and stability of multicast tree. In the condition of heterogeneous bandwidth of end host, the algorithm works efficiently, distributes server's resource in different data streaming well, offers clients links with great service quality and increase the use ratio of server system resource.